Searcharxiv⌕ Search

SEARCH · Searcharxiv

Search Searcharxiv

Search indexed arXiv papers on artificial intelligence, large language models, computer vision and robotics. Read source abstracts and follow links to arXiv.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 325 records · Page 18Linked to original sources
arXiv2026

Brain-to-Language Decoding: Tasks, Signals, Methods, Evaluation, Practical Use and Beyond

Brain-to-language decoding translates neural activity associated with language production, internal speech and perception into linguistic or expressive outputs. It offers a route to restoring communication after speech loss and a means of studying how the brain represents language. Advances in neural recording and representation learning have expanded the field from constrained recognition and acoustic reconstruction to text generation, streaming personalised speech and facial animation. This survey synthesises these developments across invasive and non-invasive measurements, drawing on a search without a lower year limit and source-led updates through September 2026. We connect Articulated, Inner and Perceived tasks to the neural populations they engage, the representations available to decoders and the outputs those representations can support. We examine model development, public resources and the evolution of evaluation, and compare published performance and communication costs within their reported protocols. The synthesis identifies complementary routes to progress: phonetic, acoustic and semantic targets preserve different aspects of a message; shared representations support reuse across recording conditions and tasks; and online communication increasingly depends on calibration, feedback and user control alongside decoding accuracy. Shared benchmarks enable algorithmic comparisons, while longitudinal studies reveal the demands of sustained use. We discuss these developments and their remaining limitations, then outline a prospective five-level trajectory from commands and language to meaning, scenarios and bidirectional cognitive exchange

cs.CL↗
arXiv2026

Mixing profile for Glauber dynamics of the discrete Gaussian Free Field starting from super-harmonic functions

We study the convergence rate of the heat-bath Glauber dynamics for the Discrete Gaussian Free Field on arbitrary connected finite graphs. We show that, when starting from super-harmonic initial conditions, the evolution enjoys a strong form of monotonicity. This allows us to get a sharp mixing profile as the size of the graphs diverges. More precisely, we show that mixing occurs at time $\frac{1}{2λ}\log(\mathcal{E})$ with window $\mathcal{O}(1/λ)$, where $λ$ is the spectral gap of the graph Laplacian and $\mathcal{E}$ is the energy of the super-harmonic initial condition. This result holds for arbitrary graphs that do not exhibit extreme connectivity properties (one way or the other). In particular, it holds for finite boxes of the grid $\mathbb{Z}^d$, in dimension $d\geq 3$.

math.PR↗
arXiv2026

Consequential Behaviour and Representational Fairness in the Validation of Synthetic Research

Researchers in industry and academia use synthetic survey respondents powered by large language models as substitutes for human samples. These synthetic populations require validation against real-world data, so researchers often address them using ad hoc comparisons with human surveys. Inspired by the intention-behaviour gap in behavioural science, we argue that these validations test the wrong thing for most applied cases where decision makers commission synthetic research to anticipate consequential behaviour. To address this problem, we propose a validation framework with two requirements. First, every validity claim must state its level of correspondence with human data: does the sample predict what the represented people do, which of four diagnostics (location, dispersion, response process and structure) does the validation address, and does the validation compare against experimental effects? Second, researchers must report validity claims for subgroups, since these groups are often the most affected by consequential decisions and aggregate accuracy hides their misrepresentation. Our validation framework operationalises three justice dimensions (distributional, procedural, and recognition) as measurable quantities and defines within-persona counterfactual experiments as a validation requirement. We then apply the framework to electric vehicle charging tariffs, before closing with a reporting checklist that researchers can use to make convincing validity claims.

cs.CL↗
arXiv2026

Gaussian polytopes with large Banach-Mazur distance to the cross-polytope

Let $B_1^n$ be the standard cross-polytope in $\mathbb R^n$, let $g_1,\ldots,g_m$ be independent standard Gaussian vectors in $\mathbb R^n$, and set $G_m=\operatorname{conv}{\pm g_1,\ldots,\pm g_m}$. For $m=n^3$ it is proved that $$ \mathbb P\left\{d_{\mathrm{BM}}(G_m,B_1^n)\geqslant c n^{5/8}(\ln n)^{-5/8}\right\}\geqslant 1-\frac2n $$ for a suitable absolute constant $c>0$. This independently improves the polynomial exponent $4/7$ in Friedland's preceding work. Independent concurrent work of Friedland, which appeared after completion of the present manuscript, obtains the same polynomial exponent with the stronger logarithmic factor $(\ln n)^{-1/4}$ by a different argument. The proof uses Friedland's discretization and conditioning argument together with the $K/U$ decomposition. A selected family of $K$ vectors is suppressed and the remaining $K$ vectors are quotiented out. In the resulting quotient simultaneous bounds are proved for every top-dimensional exterior product formed from the suppressed $K$ vectors and the $U$ vectors. A Dvoretzky-Rogers selection after L"owner normalization converts these determinant estimates into a bound for the minimum volume ellipsoid of the whole projected polytope and Maurey's empirical method then gives the required Gaussian measure estimate.

math.FA↗
arXiv2026

A Law of Fractional Logarithm for Nested Complex Sample Covariance Matrices

We prove a law of fractional logarithm for the largest eigenvalue along a northwest-nested path of complex sample covariance matrices from one infinite array. The entries are independent and centered, with unit variance, vanishing complex second moment, and uniformly bounded moments of every fixed order. The row dimension is nondecreasing, has bounded increments, and has a positive limiting aspect ratio. After finite-size edge centering and scaling, the almost-sure limsup on the $(\log N)^{2/3}$ scale is $(1/4)^{2/3}$, and the liminf on the $(\log N)^{1/3}$ scale is $-4^{1/3}$. The corresponding cluster sets in $\mathbb{R}$ are $[0,(1/4)^{2/3}]$ and $[-4^{1/3},\infty)$. The proof compares the Laplace transform of a single smoothed count over a growing grid of full nested matrices with its Gaussian counterpart. Gaussian count concentration gives block occurrences with probability tending to one. Dyadic tail bounds yield the endpoints, and deterministic interpolation gives the cluster sets.

math.PR↗
arXiv2026

Dense Interprocedural Dominance in Acyclic Graphs: Context Bounds and Compact Queries

Interprocedural dominance asks which program points every matched call-and-return execution must pass on its way to a target. An analysis that writes out immediate dominance pays for the transitive reduction of this order, the minimal dominator graph. Measured instances of that graph look near-linear, with edge-to-node ratios around 1.1--1.26~\cite{DeSutter2007}, but no program property is known to force this: it is unclear which structural restrictions imply sparse output, and whether dense output also rules out cheap queries. This paper answers both through calling contexts. A static node stands for the group of its reachable configurations; write $κ$ for the largest group size and $α$ for how many nodes have more than one configuration. One context per node keeps the tree with $n-1$ edges. Two already allow $Θ(n^2)$ cover edges in a sparse, loop-free, recursion-free ICFG with total degree three, constant-size procedures, and logarithmic stack depth, while the reachable configuration graph itself stays linear, so the blowup comes from grouping, not from many configurations. Three keep $Θ(n^2)$ covers even when the explicit call/return graph is acyclic: all target-reaching runs visit the mandatory procedures in the same order, and extra runs that stop short of any target remove the comparabilities that would otherwise kill the covers. Complementing the lower bounds, $α$ ambiguous nodes admit at most $\min\{\floor{n^2/4},(α+1)(n-1)\}$ cover edges, with $Ω(αn)$ attainable at stack depth one. And for every ICFG with at most two contexts per node, deleting a context group reduces dominance to dual-failure reachability, giving $O(n)$-word, $O(1)$-query representation after polynomial preprocessing. Explicit cover size, context ambiguity, and query-space cost are three separate quantities.

cs.DS↗
arXiv2026

Bergman's amalgamation problem over the circle

We prove that $S^1$, ${\rm SU}(2)$, and ${\rm SO}(3)$ are compact amalgamation bases: for each of these groups, any two compact groups with specified embedded copies of it embed in a compact group in which those copies agree. For compact Lie groups, the amalgam can be chosen to be a finite-dimensional unitary group. This research was conducted with substantial AI assistance, including in the development of mathematical constructions and proof arguments.

math.GR↗
arXiv2026

A Tight Cycle-Cover Inequality for Shortest Common Superstring

In the Shortest Common Superstring problem (SCS), one is given a finite set of strings and is asked to find a shortest string containing every input string as a substring. Its best known approximation ratio is $2.466$, whereas the currently strongest upper bound on the approximation guarantee of the maximum-overlap greedy algorithm is $3.396$ (Englert, Matsakis, and Vesel{ý}, 2023), though it is conjectured to be $2$. We improve both approximation guarantees: SCS admits a $\frac{7}{3}$ approximation and the approximation guarantee of the greedy algorithm is at most $3$. The main technical ingredient of our proof is a certain inequality for optimum cycle covers of an overlap graph associated with the input strings. Every previous improvement of greedy's worst-case guarantee and the two recent record guarantees for general SCS are driven by it. We improve this inequality by pushing it to its limit: for a particular coefficient of this inequality, we show a new upper bound and prove that it cannot be improved further.

cs.DS↗
arXiv2026

MIDIBack: Harmony-Aware Singing Pitch Correction via Joint Vocal-Accompaniment Symbolic Modeling

Automatic pitch correction (APC) requires distinguishing the unintended intonation errors from expressive pitch variation. Existing systems either lack explicit harmonic modeling, as vocal-only methods do, or do not directly use the note-level polyphonic context. Therefore, we propose MIDIBack, a note-level APC framework that jointly models the vocal and accompaniment events in a shared OctupleMIDI sequence. We evaluate MIDIBack under 6 note corruption regimes, including global outshift, learned note-dependent detuning, uniform perturbations, and their combinations. The resulting model achieves 78.6% overall raw pitch accuracy (RPA), and 81.5% under combined global outshift and learned detuning. Removing the accompaniment conditioning reduces RPA from 81.5% to 35.8% in outshift, showing the effectiveness of accompaniment context. Case studies on accompaniment modulation further illustrate that vocal note predictions

cs.SD↗
arXiv2026

Lettericity Is NP-Complete

The lettericity of a graph $G$ is the smallest size of a set $Σ$ such that there exist $w_1, \ldots, w_{|V(G)|} \in Σ$ and a decoder $D \subseteq Σ^2$ for which $G$ is isomorphic to the letter graph $(\{1, \ldots, |V(G)|\}, \{ij : 1 \le i < j \le |V(G)|, w_iw_j \in D\})$. It took around two decades of the study of lettericity for, in the simpler case of paths, a closed-form expression for its lettericity to be derived; this suggests that the question of whether the lettericity of an arbitrary graph can be computed in polynomial time is nontrivial. Indeed, this question has been raised repeatedly as an open problem in recent literature. We solve this problem by showing that the lettericity problem on arbitrary graphs is NP-complete (Theorem 10). We also prove that the coloring extension problem (the same problem as lettericity, with the added condition that if $f$ is the isomorphism mapping from $G$ to the letter graph, $w_{f(v)} = χ(v)$ for a given coloring $χ$ of $G$) is NP-complete (Theorem 12). We also resolve the open problem of classifying the complexity of the word extension problem, which is the same problem as lettericity except that the $w_i$ are fixed; we show it to be NP-complete (Theorem 13), which, in tandem with our NP-completeness result for coloring extension, contrasts with the known result that when the constraint of the coloring extension problem and the constraint of the word extension problem are both applied to lettericity, lettericity can be decided in polynomial time. Additionally, we use the reduction in the NP-completeness proof to show that unless the Exponential Time Hypothesis is false, there cannot exist a deterministic algorithm to decide whether the lettericity of an $n$-vertex graph is at most $k$ in time $2^{o(n)}$, even when $n = 6k$ (Theorem 11).

cs.CC↗
arXiv2026

Strangeness Production in Heavy-Ion Collisions: Color Ropes or Hydrodynamic Evolution?

We investigate strangeness production and transverse dynamics in heavy-ion collisions at $\sqrt{s_{\mathrm{NN}}}\approx 2.5-20~\mathrm{GeV}$ using the transport approach SMASH (Simulating Many Accelerated Strongly-interacting Hadrons), its extension with rope hadronization, and the SMASH+vHLLE hybrid approach. Results from the Pythia-based heavy-ion model Angantyr, with and without rope hadronization, are included for comparison. We study midrapidity particle yields and average transverse masses as functions of the number of wounded nucleons, as well as their energy dependence. For the $K^+/π^+$ ratio, SMASH+vHLLE overpredicts strangeness production at low energies but describes the higher-energy behavior reasonably well. SMASH+Ropes reproduces the ratio up to $\sqrt{s_{\mathrm{NN}}}\sim 10~\mathrm{GeV}$ but does not capture the turnover at higher energies. In contrast, the transverse-mass observables favor the hybrid approach, while the non-thermal models considered here do not generate sufficient collective transverse expansion. These results show that strangeness enhancement alone does not uniquely distinguish microscopic string interactions from a locally equilibrated medium. Simultaneously constraining strangeness production and transverse dynamics is therefore essential for disentangling thermal and non-thermal mechanisms in heavy-ion collisions.

nucl-th↗
arXiv2026

LAYERSCOPE: A Layerwise Characterization of Video and Multimodal Learned Representations

We propose LAYERSCOPE, a label-free, layerwise framework that aims to characterize a model's learned representations in video and multimodal settings. Evaluating downstream performance using representations from final or intermediate layers typically requires large amounts of labeled data, repeated task-specific evaluations, and substantial computation. To address these limitations, LAYERSCOPE uses local, global, distributional, and correspondence-based geometric metrics to compare layerwise representation structure within and across models without requiring task-specific labels. We evaluate seven architecturally diverse models across video and multimodal classification, clustering, and text-to-video retrieval tasks from MVEB/MVEB+. We find that intermediate-layer representations can outperform final-layer and model-default outputs. We also find that no single geometric metric consistently predicts downstream performance, but note that distinct layerwise geometric signatures emerge across model families. LID shows task-dependent relationships with performance, while RankMe provides the strongest measure for classification and clustering, but is not a universal layer selector. We also find that pairing-aware metrics explain retrieval better than distributional distances alone. LAYERSCOPE therefore offers a framework for comparing representations across models and layers, enabling a more systematic evaluation in video and multimodal settings.

cs.LG↗
arXiv2026

Distillation for Efficient Multitask Manipulation Policies via Conditional Flow Matching

Advances in generative modeling have recently been extensively employed in robotics for policy learning. In particular, Conditional Flow Matching (CFM) trained with expert demonstrations has been shown to outperform existing methods on robot manipulation benchmarks. While prior work has mainly focused on single-task settings, we study the problem from a multi-task perspective, as training independent models for each task is computationally expensive. Multi-Task policy learning comes with its own set of challenges, as naively training on a concatenated dataset of demonstrations would either require increased model capacity to accommodate the added complexity or result in drops in performance. We propose to distill knowledge from single-task CFM experts into a shared multi-task policy by transferring their learned velocity fields. We combine this distillation signal with the original CFM objective to retain fidelity to the demonstrations. Experiments on RLBench show that our approach improves multi-task policy performance over naive training while maintaining a fixed model size.

cs.RO↗
arXiv2026

UNITE-AUDIO: Joint Learning of Continuous Tokenization and Latent Flow Matching for Text-to-Audio Generation

Text-to-audio (TTA) generation aims to synthesize realistic audio that faithfully reflects natural-language descriptions. Most TTA systems adopt a two-stage latent paradigm: an audio tokenizer is optimized for reconstruction and then frozen, after which a generative model is trained in the resulting latent space. However, reconstruction-oriented representations may be suboptimal for generation, motivating joint representation and generative learning. To this end, we introduce Unite-Audio, to our knowledge, is the first to jointly learn continuous audio representations and latent flow matching for TTA. By coupling reconstruction with self-supervised generative prediction, Unite-Audio allows the generative objective to directly shape the latent space rather than treating it as a fixed intermediate representation. We further employ Flow-GRPO post-training to improve text-conditioned generation. Experiments show competitive TTA performance with a compact latent flow model, while ablation studies confirm the benefit of jointly learning the audio representation and generative model. Audio samples are available at https://runwushi.github.io/Unite-Audio.

eess.AS↗
arXiv2026

Learning the Cost of Reliable Inference

Benchmarking and routing platforms increasingly act as intermediaries connecting large language model providers with end-users. However, providers on these platforms typically use a fixed price per token, preventing users from achieving the most competitive price for their tasks. In this work, we design a procurement platform where token prices for each task are driven by provider competition, enabling users to secure competitive pricing for guaranteed quality levels. To this end, the platform sequentially routes queries via a reverse second-price auction that incentivizes model providers to truthfully bid their best estimate of the average cost to serve a user's query. As it routes queries, the platform learns the quality offered by each provider and progressively routes queries to the most cost-competitive provider among those meeting a desired quality threshold. To validate our design, we conduct experiments with multiple LLMs from the Llama and Qwen families on popular mathematical reasoning and question-answering benchmarks. The results show that the pricing margin of the most cost-competitive provider on our platform varies significantly---from $10\%$ to $71\%$---depending on the task and quality threshold. This suggests a substantial inefficiency in the current fixed-price market, and it demonstrates that our platform may enable users to capture maximum savings whenever competitive market conditions permit.

cs.AI↗
arXiv2026

"You Can't Just Automate It": Negotiating and Sustaining a "Good" Family Life Through Energy Practices

This study examines how Taiwanese parent-child families negotiate a "good" family life through everyday energy use and imagine future smart homes that support it. We conducted in-home interviews and co-design sessions with 21 families, including 46 parents and children. We found that families pursued a good life through energy practices shaped by thrift, comfort, care, safety, and enjoyment. These arrangements were continually adapted and responded to changing bodies, schedules, people, and infrastructures. This adaptive work was unevenly distributed, which in turn shaped different smart-home imaginaries. Drawing on the lens of Nearby and adversarial design, we conceptualize adaptation as situated sociotechnical work through which families continually rework energy arrangements. We further distinguish collective goods from plural and contestable goods to show why family IoT must support shared values while preserving opportunities to question and revise household arrangements. We offer theoretical and design directions for more adaptive, participatory, and contestable family IoT.

cs.HC↗
arXiv2026

Human-AI-Powered Hypothesis Testing: Cost-Aware Selective AI Scoring and Sequential Human Escalation

Large language models are increasingly used as inexpensive judges to evaluate outputs, label data, and assess whether a system meets a desired quality standard. Yet using AI judgments for formal statistical inference is fundamentally different from simply treating them as ground-truth labels: AI evaluations can be biased or noisy, and rigorous hypothesis testing requires explicit control of type-I and type-II errors. We study how to use AI judgments, together with selective human verification, to conduct a valid hypothesis test at minimum cost. We consider a population of items with hidden binary labels. After choosing a fixed pool of items, the decision maker can selectively query AI, send an item directly to a human, escalate an AI-scored item to a human after observing the AI report, or stop once sufficient evidence has accumulated. We derive an information-theoretic lower bound that captures the minimum cost of achieving prescribed testing errors and characterizes the value of AI information and human verification through a report-dependent information frontier. Motivated by this characterization, we develop SCALE, a sequential cost-aware policy that combines selective AI scoring with adaptive human escalation. SCALE is valid at finite sample sizes and matches the lower bound to first order as the target error probabilities vanish. We further extend the framework to an unknown AI-output model using paired AI-human pilot data. Numerically, SCALE approaches Human-only or AI-only testing when one source clearly dominates, while achieving its largest savings when inexpensive AI judgments and selective human verification are both valuable.

cs.AI↗
arXiv2026

Multimodal Routing and Region Refinement for Language-Guided Medical Image Segmentation

Textual descriptions can reduce ambiguity in medical image segmentation by specifying the finding and location to be delineated. Existing text-guided methods mainly improve where image and language features interact but generally retain a single learned update pathway across all image-text pairs. We propose MRSeg, a parameter-efficient framework that uses each image-text pair to route the adaptation of visual and textual features before dense prediction. Frozen ConvNeXt-Tiny and PubMedBERT encoders provide multiscale visual features and clinical text tokens. A joint router uses the deepest visual feature and pooled text to predict a sparse mixture over low-rank adapter bases. The resulting route is shared across separate adapter banks for two visual scales and text, coordinating their adaptation while keeping the feature-specific parameters separate. Region Bridge uses text-derived queries to aggregate dense visual tokens into latent regions, refines these regions through self-attention and text cross-attention, and redistributes the refined information back to the feature maps. Finally, a multiscale decoder combines refined semantic features with shallow image evidence. On QaTa-COV19 and MosMedData+, MRSeg achieves 90.90/83.32 and 81.53/68.82 Dice/mIoU, respectively, with 7.11M trainable parameters and 7.60 GFLOPs. Code: https://github.com/maklachur/MRSeg.

cs.CV↗